How PCR amplification affects rna-seq data

Question

[Javad]

Dear all,

In certain RNA-seq protocols, RNA is amplified via PCR. For example SMART-seq or some other single cell RNA-seq methods.

Whereas this amplification is required due to the low amount of initial RNA, it can also introduce bias. For instance PCR favors highly abundant transcripts and they are highly amplified. On the other hand, low abundant transcripts are only moderately amplified.

This can distort the whole quantification process and the number of reads after RNAseq won’t reflect the abundance of transcripts before amplification. In principle, the effect of amplification can be huge and can potentially distort the whole experiment. I have difficulty to understand how methods like smart-seq (or others with the same principle) work.

Can someone explain how this can be explained?

Of course if unique molecular identifiers (UMIs) are used, the effect of PCR can be regressed out. What I mean with this question is about protocols that do not use UMIs.

Thanks in advance

[igor]

There are some previous discussions on the topic:

• Duplicated Reads In Rna-Seq Experiment
• How detrimental are duplicate reads in RNAseq experiments?

[dsull]

Good reading on the subject about the biases of non-UMI protocols: https://www.nature.com/articles/s41587-020-00810-6

Unfortunately, there's a tradeoff between 5'-end/3'-end bias (UMI technologies) and amplification bias (non-UMI technologies) and there's currently no solution.